Sorry but you're absolutely wrong! While it's true that FaceID uses neural nets to train on your face, it does this ONCE. This means that FaceID won't improve each time it sees your mug.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Face ID Appears to Fail at Telling Apart Brothers Who Aren't Twins in New Videos [Updated]
- Thread starter MacRumors
- Start date
- Sort by reaction score
Sorry but you're absolutely wrong! While it's true that FaceID uses neural nets to train on your face, it does this ONCE. This means that FaceID won't improve each time it sees your mug.
Haha yeah that’s what Apple said. It’s lie , I have seen at least 15 cases by now and Apple didn’t ship 15 million phones i guess. What a big lie!
Wow. You clearly don't understand what probabilities represent.
You my friend, are VERY wrong. Apple has clearly stated this several times. I suggest you watch the keynote.
No. You are wrong. Apple states quite clearly that it constantly learns, adapts to changes in facial features over time, and pays particular attention when a match fails and is followed by entering of the correct passcode.
I didn’t read that...there is a lot to read. So we were both right.
I’m glad to see someone else has a brain.
[doublepost=1509936279][/doublepost]
It does work...technology is always evolving. It’s never a finished product. If you don’t like it, use Android which throws out tech for authentication that they don’t even back with their own Samsung Pay.
Apple has confidence in this.
The “failures” are based on non real world scenarios with people purposely trying to fool it over and over in its early learning stages of adaptation while mitigating failures by authenticating with a passcode. They are teaching it to recognize multiple users. This will only happen when people are not using it as intended. Wake me up when someone picks up a random phone and unlocks it.
You say it should never fail, but this isn’t magic. It’s based on user input which if done properly, will not be fooled in this manner.
Last edited:
It’s actually pretty great unless you have an identical twin or you give your passcode to your sibling.
Sure. The stated spec is a 1 in a million chance of a RANDOM person being able to unlock the phone.
What you cited is a number of reported attempts, some intentionally cajoled. So that's not the same numerator. Then you divided by an assumed number of phones sold. That's not the right denominator either.
samsung pay can be unlocked via passcode, fingerprint and iris scanner.
[doublepost=1509944277][/doublepost]
so what you're basically saying is that if the people chosen aren't random then theres a much bigger chance than 1 in a million for them to unlock your phone? lol yeah that makes it so much better.....
yeah even samsung said that wasn't mean to be used for real security. but yet here is apple using facial recognition for apple pay lol.....
Exactly, so entering the password after a fail, tells FaceID it should have recognized the face.
So they "trained" FacedID to recognize both.
How is this different than training it to read a fingerprint from someone else?
Don't train it to recognize your siblings or anyone else. DUH!
That is what Apple said during the keynote.
Plus, you have no clue how much failure video are authentic. How much show the registration ? For how much video you are sure the second people had not enter the code ?
With only positive reinforcement how is the algorithm going to know his brother is not just another "look" ?
If you lose your phone the same day you get it, you deserve to be screwed lol
[doublepost=1509979870][/doublepost]
Someone else in the board already said that they were able to do it with a similar relative. So this whole article is pointless.
You don't think it's because Apple is using completely different technology that actually can be secure enough for mobile payment?
[doublepost=1509981363][/doublepost]
No one. It's terrible.